kolibrios/programs/demos/view3ds/3dmath.inc
IgorA ec81f654c0 view3ds 076 by macgub
git-svn-id: svn://kolibrios.org@9512 a494cfbc-eb01-0410-851d-a64ba20cac60
2021-12-28 08:18:43 +00:00

730 lines
15 KiB
PHP

x3d equ 0
y3d equ 2
z3d equ 4
vec_x equ 0
vec_y equ 4
vec_z equ 8
if 0 ; Ext >= SSE3
calc_bounding_box:
; in:
; xmm0 - normal vector of ray
; xmm1 - light origin
; out:
; eax - axis aligned bounding boxes bit mask
.rmx equ [ebp-36]
.nray equ [ebp-64]
.origin equ [ebp-80]
.dirfrac equ [ebp-96]
.nrayr equ [ebp-112]
.originr equ [ebp-128]
.tmin equ [ebp-132]
.tmax equ [ebp-136]
push ebp
mov ebp,esp
and ebp,-16
sub esp,160
movss xmm5,[rsscale]
shufps xmm5,xmm1,0
movd xmm2,[vect_x]
punpcklwd xmm2,[the_zero]
cvtdq2ps xmm2,xmm2
subps xmm1,xmm2
movaps .origin,xmm1
mulps xmm0,xmm5
movaps .nray,xmm0
mov esi,matrix
lea edi,.rmx
call reverse_mx_3x3
; in: esi - ptr to points(normals], each point(normal) coeficient as dword
; edi - ptr to rotated points(normals)
; ebx - ptr to 3x3 (9 dwords, 36 bytes) rotation matrix
; ecx - number of points(normals)
; reverse transform
lea esi,.nray
lea edi,.nrayr
lea ebx,.rmx
mov ecx,1
call rotary
lea esi,.origin
lea edi,.originr
lea ebx,.rmx
mov ecx,1
call rotary
xor ecx,ecx
mov ebx,aabb1
xor eax,eax
rcpps xmm7,.nrayr
movaps .dirfrac,xmm7
.nx_aabb:
movaps xmm5,[ebx]
movaps xmm6,[ebx]
minps xmm5,[the_zero]
maxps xmm6,[the_zero]
; xmm5 - lb corner of AABB with minimal coordinates
; xmm6 - rt cor. of AABB wit maximum coords
subps xmm5,.originr
subps xmm6,.originr
mulps xmm5,.dirfrac ; xmm5 - tx1, ty1
mulps xmm6,.dirfrac ; xmm6 - tx2, ty2
movaps xmm1,xmm6
movaps xmm2,xmm6
minps xmm1,xmm5
maxps xmm2,xmm5
movaps xmm5,xmm1
movaps xmm6,xmm2
shufps xmm5,xmm5,11100001b
shufps xmm6,xmm6,11100001b
maxss xmm1,xmm5 ;t min
minss xmm2,xmm6 ;t max
comiss xmm2,xmm1
jb .no_inter
.yes:
bts eax,ecx
.no_inter:
add ebx,16
inc ecx
cmp ecx,8
jne .nx_aabb
; out: eax - bit mask
add esp,160
pop ebp
ret
end if
reverse_mx_3x3:
; esi - source matrix
; edi - desired reversed matrix
push ebp
mov ebp,esp
sub esp,4
.det equ ebp-4
fninit
fld dword[esi]
fmul dword[esi+16]
fmul dword[esi+32]
fld dword[esi+12]
fmul dword[esi+28]
fmul dword[esi+8]
faddp
fld dword[esi+24]
fmul dword[esi+4]
fmul dword[esi+20]
faddp
fld dword[esi]
fmul dword[esi+28]
fmul dword[esi+20]
fchs
faddp
fld dword[esi+24]
fmul dword[esi+16]
fmul dword[esi+8]
fchs
faddp
fld dword[esi+12]
fmul dword[esi+4]
fmul dword[esi+32]
fchs
faddp
fstp dword[.det]
cmp dword[.det],0
jne @f
int3
@@:
; fld1
; fdiv dword[.det]
; fstp dword[.det]
fld dword[esi+16]
fmul dword[esi+32]
fld dword[esi+20]
fmul dword[esi+28]
fchs
faddp
fdiv dword[.det]
fstp dword[edi]
fld dword[esi+8]
fmul dword[esi+28]
fld dword[esi+4]
fmul dword[esi+32]
fchs
faddp
fdiv dword[.det]
fstp dword[edi+4]
fld dword[esi+4]
fmul dword[esi+20]
fld dword[esi+8]
fmul dword[esi+16]
fchs
faddp
fdiv dword[.det]
fstp dword[edi+8]
fld dword[esi+20]
fmul dword[esi+24]
fld dword[esi+12]
fmul dword[esi+32]
fchs
faddp
fdiv dword[.det]
fstp dword[edi+12]
fld dword[esi]
fmul dword[esi+32]
fld dword[esi+8]
fmul dword[esi+24]
fchs
faddp
fdiv dword[.det]
fstp dword[edi+16]
fld dword[esi+8]
fmul dword[esi+12]
fld dword[esi]
fmul dword[esi+20]
fchs
faddp
fdiv dword[.det]
fstp dword[edi+20]
fld dword[esi+12]
fmul dword[esi+28]
fld dword[esi+16]
fmul dword[esi+24]
fchs
faddp
fdiv dword[.det]
fstp dword[edi+24]
fld dword[esi+4]
fmul dword[esi+24]
fld dword[esi]
fmul dword[esi+28]
fchs
faddp
fdiv dword[.det]
fstp dword[edi+28]
fld dword[esi]
fmul dword[esi+16]
fld dword[esi+4]
fmul dword[esi+12]
fchs
faddp
fdiv dword[.det]
fstp dword[edi+32]
mov esp,ebp
pop ebp
ret
; 3d point - triple integer word coordinate
; vector - triple float dword coordinate
;----------------------in: --------------------------------
;------------------------ esi - pointer to 1st 3d point ---
;------------------------ edi - pointer to 2nd 3d point ---
;------------------------ ebx - pointer to result vector --
;---------------------- out : none ------------------------
make_vector_r:
if Ext < SSE
fninit
fld dword[edi] ;edi+x3d
fsub dword[esi] ;esi+x3d
fstp dword[ebx+vec_x]
fld dword[edi+4]
fsub dword[esi+4]
fstp dword[ebx+vec_y]
fld dword[edi+8]
fsub dword[esi+8]
fstp dword[ebx+vec_z]
else
movups xmm0,[esi]
movups xmm1,[edi]
subps xmm1,xmm0
movlps [ebx],xmm1
movhlps xmm1,xmm1
movss [ebx+8],xmm1
end if
ret
;---------------------- in: -------------------------------
;--------------------------- esi - pointer to 1st vector --
;--------------------------- edi - pointer to 2nd vector --
;--------------------------- ebx - pointer to result vector
;---------------------- out : none
cross_product:
fninit
fld dword [esi+vec_y]
fmul dword [edi+vec_z]
fld dword [esi+vec_z]
fmul dword [edi+vec_y]
fsubp ;st1 ,st
fstp dword [ebx+vec_x]
fld dword [esi+vec_z]
fmul dword [edi+vec_x]
fld dword [esi+vec_x]
fmul dword [edi+vec_z]
fsubp ;st1 ,st
fstp dword [ebx+vec_y]
fld dword [esi+vec_x]
fmul dword [edi+vec_y]
fld dword [esi+vec_y]
fmul dword [edi+vec_x]
fsubp ;st1 ,st
fstp dword [ebx+vec_z]
ret
cross_aligned:
; params as above cross_p
movaps xmm0,[esi]
movaps xmm1,[esi]
movaps xmm2,[edi]
movaps xmm3,[edi]
shufps xmm0,xmm0,00001001b
shufps xmm1,xmm1,00010010b
shufps xmm2,xmm2,00010010b
shufps xmm3,xmm3,00001001b
mulps xmm0,xmm2
mulps xmm1,xmm3
subps xmm0,xmm1
movaps [ebx],xmm0
ret
;----------------------- in: ------------------------------
;---------------------------- edi - pointer to vector -----
;----------------------- out : none
normalize_vector:
if Ext >= SSE2
movups xmm0,[edi]
andps xmm0,[zero_hgst_dd]
movups xmm1,xmm0
mulps xmm0,xmm0
movhlps xmm2,xmm0
addps xmm0,xmm2
movaps xmm2,xmm0
shufps xmm2,xmm2,11100101b
addps xmm0,xmm2
shufps xmm0,xmm0,0
; haddps xmm0,xmm0
; haddps xmm0,xmm0
rsqrtps xmm0,xmm0
mulps xmm0,xmm1
movlps [edi],xmm0
movhlps xmm0,xmm0
movss [edi+8],xmm0
else
fninit
fld dword [edi+vec_x]
fmul st, st
fld dword [edi+vec_y]
fmul st, st
fld dword [edi+vec_z]
fmul st, st
faddp st1, st
faddp st1, st
fsqrt
ftst
fstsw ax
sahf
jnz @f
fst dword [edi+vec_x]
fst dword [edi+vec_y]
fstp dword [edi+vec_z]
ret
@@:
fld st
fld st
fdivr dword [edi+vec_x]
fstp dword [edi+vec_x]
fdivr dword [edi+vec_y]
fstp dword [edi+vec_y]
fdivr dword [edi+vec_z]
fstp dword [edi+vec_z]
end if
ret
;------------------in: -------------------------
;------------------ esi - pointer to 1st vector
;------------------ edi - pointer to 2nd vector
;------------------out: ------------------------
;------------------ st0 - dot-product
dot_product:
fninit
;if Ext >=SSE3
; movups xmm0,[esi]
; movups xmm1,[edi]
; andps xmm0,[zero_hgst_dd]
; mulps xmm0,xmm1
; haddps xmm0,xmm0
; haddps xmm0,xmm0
; movss [esp-4],xmm0
; fld dword[esp-4]
;else
fld dword [esi+vec_x]
fmul dword [edi+vec_x]
fld dword [esi+vec_y]
fmul dword [edi+vec_y]
fld dword [esi+vec_z]
fmul dword [edi+vec_z]
faddp
faddp
;end if
ret
; DOS version Coded by Mikolaj Felix aka Majuma
; mfelix@polbox.com
; www.majuma.xt.pl
; into FASM translation by Macgub
init_sincos_tab:
.counter equ dword [ebp-4] ; cur angle
push ebp
mov ebp,esp
xor eax,eax
push eax ; init .counter
mov edi,cos_tab
mov esi,sin_tab
mov ecx,256
fninit
fld .counter
@@:
fld st
fsincos
fstp dword [edi]
fstp dword [esi]
; fadd [piD180]
fadd [piD128]
add esi,4
add edi,4
loop @b
ffree st
mov esp,ebp
pop ebp
ret
;------
; esi - offset (pointer) to angles, edi offset to 3x3 matrix
make_rotation_matrix:
.sinx equ dword[ebp-4]
.cosx equ dword[ebp-8]
.siny equ dword[ebp-12]
.cosy equ dword[ebp-16]
.sinz equ dword[ebp-20]
.cosz equ dword[ebp-24]
push ebp
mov ebp,esp
sub esp,24
movzx ebx,word[esi]
shl ebx,2
mov eax,dword[sin_tab+ebx]
mov .sinx,eax
mov edx,dword[cos_tab+ebx]
mov .cosx,edx
movzx ebx,word[esi+2]
shl ebx,2
mov eax,dword[sin_tab+ebx]
mov .siny,eax
mov edx,dword[cos_tab+ebx]
mov .cosy,edx
movzx ebx,word[esi+4]
shl ebx,2
mov eax,dword[sin_tab+ebx]
mov .sinz,eax
mov edx,dword[cos_tab+ebx]
mov .cosz,edx
fninit
fld .cosy
fmul .cosz
fstp dword[edi]
fld .sinx
fmul .siny
fmul .cosz
fld .cosx
fmul .sinz
fchs
faddp
fstp dword[edi+12]
fld .cosx
fmul .siny
fmul .cosz
fld .sinx
fmul .sinz
faddp
fstp dword[edi+24]
fld .cosy
fmul .sinz
fstp dword[edi+4]
fld .sinx
fmul .siny
fmul .sinz
fld .cosx
fmul .cosz
faddp
fstp dword[edi+16]
fld .cosx
fmul .siny
fmul .sinz
fld .sinx
fchs
fmul .cosz
faddp
fstp dword[edi+28]
fld .siny
fchs
fstp dword[edi+8]
fld .cosy
fmul .sinx
fstp dword[edi+20]
fld .cosx
fmul .cosy
fstp dword[edi+32]
mov esp,ebp
pop ebp
ret
;---------------------
; in: esi - ptr to points(normals], each point(normal) coeficient as dword
; edi - ptr to rotated points(normals)
; ebx - ptr to 3x3 (9 dwords, 36 bytes) rotation matrix
; ecx - number of points(normals)
rotary:
if Ext<SSE
fninit
.again:
fld dword[esi]
fmul dword[ebx]
fld dword[esi+4]
fmul dword[ebx+12]
faddp
fld dword[esi+8]
fmul dword[ebx+24]
faddp
fstp dword[edi]
fld dword[esi+4]
fmul dword[ebx+16]
fld dword[esi]
fmul dword[ebx+4]
faddp
fld dword[esi+8]
fmul dword[ebx+28]
faddp
fstp dword[edi+4]
fld dword[esi+8]
fmul dword[ebx+32]
fld dword[esi]
fmul dword[ebx+8]
fld dword[esi+4]
fmul dword[ebx+20]
faddp
faddp
fstp dword[edi+8]
add esi,12
add edi,12
loop .again
mov [edi],dword -1
else
; Copyright (C) 1999-2001 Brian Paul
; Copyright (C) Maciej Guba
;---------------------
; in: esi - ptr to points(normals], each point(normal) coeficient as dword
; edi - ptr to rotated points(normals)
; ebx - ptr to 3x3 (9 dwords, 36 bytes) rotation matrix
; ecx - number of points(normals)
;align 32
movups xmm4,[ebx]
; lddqu xmm4,[ebx] ; I tried sse3 :D
movups xmm5,[ebx+12]
movups xmm6,[ebx+24]
;align 32
.again:
movss xmm0,dword[esi]
shufps xmm0,xmm0,0
mulps xmm0,xmm4
movss xmm1,dword[esi+4]
shufps xmm1,xmm1,0
mulps xmm1,xmm5
movss xmm2,dword[esi+8]
shufps xmm2,xmm2,0
mulps xmm2,xmm6
addps xmm0,xmm1
addps xmm0,xmm2
movups [edi],xmm0
add esi,12
add edi,12
dec ecx
jne .again
mov [edi],dword -1
end if
ret
;----------------------------------------------
; esi - pointer to 3x3 matrix
add_scale_to_matrix:
if Ext>SSE
movss xmm0,[rsscale]
shufps xmm0,xmm0,0
movups xmm1,[esi]
movups xmm2,[esi+16]
movss xmm3,[esi+32]
mulps xmm1,xmm0
mulps xmm2,xmm0
mulss xmm3,xmm0
movups [esi],xmm1
movups [esi+16],xmm2
movss [esi+32],xmm3
else
fninit
fld [rsscale]
fld dword[esi] ;-----
fmul st,st1
fstp dword[esi]
fld dword[esi+12] ; x scale
fmul st,st1
fstp dword[esi+12]
fld dword[esi+24]
fmul st,st1
fstp dword[esi+24] ;------
fld dword[esi+4] ;-----
fmul st,st1
fstp dword[esi+4]
fld dword[esi+16] ; y scale
fmul st,st1
fstp dword[esi+16]
fld dword[esi+28]
fmul st,st1
fstp dword[esi+28] ;------
fld dword[esi+8] ;-----
fmul st,st1
fstp dword[esi+8]
fld dword[esi+20] ; z scale
fmul st,st1
fstp dword[esi+20]
fld dword[esi+32]
fmulp st1,st
fstp dword[esi+32] ;------
end if
ret
;in esi - offset to 3d points (point as 3 dwords float)
; edi - offset to 2d points ( as 3 words integer)
; ecx - number of points
translate_points: ; just convert into integer; z coord still needed
if Ext < SSE
fninit
else
; movaps xmm1,[vect_x]
end if
.again:
if 0
fld dword[esi+8]
; fmul [rsscale]
fist word[edi+4]
fisub [zobs]
fchs
fld dword[esi]
; fmul [rsscale]
fisub [xobs]
fimul [zobs]
fdiv st0,st1
fiadd [xobs]
fiadd [vect_x]
fistp word[edi]
fld dword[esi+4]
; fmul [rsscale]
fisub [yobs]
fimul [zobs]
fdivrp ; st0,st1
fiadd [yobs]
fiadd [vect_y]
fistp word[edi+2]
end if
if Ext>=SSE
movups xmm0,[esi]
cvtps2dq xmm0,xmm0
packssdw xmm0,xmm0
paddw xmm0,[vect_x]
movd [edi],xmm0
; psrldq xmm0,4
; movd eax,xmm0
pextrw eax,xmm0,6
mov [edi+4],ax
else
; cvtps2dq xmm0,xmm0
; packsdw xmm0,xmm0
; movq [edi]
fld dword[esi]
fiadd word[vect_x]
fistp word[edi]
fld dword[esi+4]
fiadd [vect_y]
fistp word[edi+2]
fld dword[esi+8]
fistp word[edi+4]
end if
add esi,12
add edi,6
; dec ecx
; jnz .again
loop .again
ret